Access Type

Open Access Dissertation

Date of Award

January 2018

Degree Type


Degree Name



Industrial and Manufacturing Engineering

First Advisor

Qingyu Yang


Most customers today are pursuing engineering materials (e.g., steel) that not only can achieve their expected functions but also are highly reliable. As a result, reliability analysis of materials has been receiving increasing attention over the past few decades. Most existing studies in the reliability engineering field focus on developing model-based and data-driven approaches to analyze material reliability based on material failure data such as lifetime data and degradation data, without considering effects of material physical properties. Ignoring such effects may result in a biased estimation of material reliability, which in turn could incur higher operation or maintenance costs.

Recently, with the advancement of sensor technology more information/data concerning various physical properties of materials are accessible to reliability researchers. In this dissertation, considering the significant impacts of steel physical properties on steel failures, we propose systematic methodologies for steel reliability analysis by integrating a set of steel physical properties. Specifically, three steel properties of various scales are considered: 1) a macro-scale property called overload retardation; 2) a local-scale property called dynamic local deformation; and 3) a micro-scale property called microstructure effect. For incorporating property 1), a novel physical-statistical model is proposed based on a modification of the current Paris law. To incorporate property 2), a novel statistical model named multivariate general path model is proposed, which is a generalization of an existing univariate general path model. For the integration of property 3), a novel statistical model named distribution-based functional linear model is proposed, which is a generalization of an existing functional linear model. Theoretical property analyses and statistical inferences of these three models are intensively developed.

Various simulation studies are implemented to verify and illustrate the proposed methodologies. Multiple physical experiments are designed and conducted to demonstrate the proposed models. The results show that, through the integration of the aforementioned three steel physical properties, a significant improvement of steel reliability assessment is achieved in terms of failure prediction accuracy compared to traditional reliability studies.